Backend has to be theano?

[xuzhang5788]

Thank you for your library.
Theano is not popular anymore. Can we use it with Tensorflow backend?
In addition, I trained several NN models and saved models as .h5 files. I want to do further ensemble, such as voting, average outputs, stacking, etc. How can I use your library? Your advice is highly appreciated.

[nitbix]

Hi, sorry for the delay in replying to this. We have decided long ago that we would stop supporting the current version and rewrite toupee from the ground up in a more elegant way, based on TensorFlow 2. You can follow development in the v2 branch. With the new design it should be easy to create new ensemble methods by extending toupee.ensembles.methods.EnsembleMethod (there are even some handy hooks that you can use.